NOTES ON SCHUR FUNCTIONS

ANDERS SKOVSTED BUCH

These notes are work in progress. The goal is to provide quick proofs of some
of the main identities satisfied by Schur functions. Some alternative references are
[Mac95, Ful97].

1. Definition of Schur functions

sec:symfcn
1.1. Symmetric functions. Let X = (x1 , x2 , . . . ) and Y = (y1 , y2 , . . . ) be two
countably infinite sets of independent commuting variables. Define the double com-
plete symmetric function Sp = Sp (X; Y ) ∈ ZJX, Y K, for p ∈ Z, by the generating
series Q∞
j=1 (1 − yj t)
X
p
Sp t = Q∞ .
p i=1 (1 − xi t)

The power series Sp is homogeneous of total degree p. We have S0 = 1 and Sp = 0

for p < 0, and the functions Sp for p ≥ 1 are algebraically independent. The ring
of symmetric functions Λ is the subring of ZJX, Y K generated by the elements Sp ,
Λ = Z[S1 , S2 , S3 , . . . ] ⊂ ZJX, Y K .
Let f ∈ Λ be a symmetric function, let R be a commutative ring, and let
a = (a1 , . . . , an ) and b = (b1 , . . . , bm ) be finite sets of elements of R. We then let
f (a; b) = f (a1 , . . . , an ; b1 , . . . , bm ) = f (a1 , . . . , an , 0, 0, . . . ; b1 , . . . , bm , 0, 0, . . . ) ∈ R
denote the result of substituting xi = ai for 1 ≤ i ≤ n, xi = 0 for i > n, yj = bj for
1 ≤ j ≤ m, and yj = 0 for j > m. We will always use a semicolon to separate the
two sets of arguments.
The resulting functions f : Rn × Rm → R are super-symmetric in the following
sense. First, f (a; b) is separately symmetric in each set of arguments a and b. In
addition, f (a; b) is unchanged if 0 is added to either set of arguments, or if the same
element c ∈ R is added to both sets of arguments:
f (a; b) = f (a, 0 ; b) = f (a ; b, 0) = f (a, c ; b, c) .
This follows from the definition of the generators Sp of Λ.
The functions Sp satisfy the following identities. If the second set of arguments
is omitted, then
Sp (a) = Sp (a; 0) = hp (a1 , . . . , an )
is the complete symmetric polynomial, defined as the sum of all monomials of degree
p in a = (a1 , . . . , an ). If the first set of arguments is omitted, then
Sp (0; b) = ep (−b1 , . . . , −bm ) = (−1)p ep (b1 , . . . , bm ) ,

Date: March 4, 2022.

1
 2 ANDERS SKOVSTED BUCH

where ep (b1 , . . . , bn ) is the elementary symmetric polynomial, defined as the sum of

all square-free monomials of degree p in b = (b1 , . . . , bm ). In general, we have
X p
X
Sp (a; b) = hi (a) ej (−b) = (−1)j hp−j (a) ej (b) .
i+j=p j=0

1.2. Schur functions. Given an integer sequence λ = (λ1 , . . . , λ` ) ∈ Z` , define

the (double) Schur function Sλ ∈ Λ by
Sλ1 Sλ1 +1 Sλ1 +2 ... Sλ1 +`−1
Sλ2 −1 Sλ2 Sλ2 +1 ... Sλ2 +`−2
Sλ = det Sλi +j−i


= Sλ3 −2 Sλ3 −1 Sλ3 ... Sλ3 +`−3 .
`×` .. .. .. ..
..
. . . . .
Sλ` −`+1 Sλ` −`+2 Sλ` −`+3 ... Sλ`
The diagonal entries in the determinant are Sλ1 , Sλ2 , . . . , Sλ` , and the subscripts
increase consecutively from left to right. For example,
S3 S4 S5 S3 S4 S5
S(3,−1,2) = S−2 S−1 S0 = 0 0 1 = S4 − S1 S3 = −S(3,1) .
S0 S1 S2 1 S1 S2
The element Sλ ∈ Λ is homogeneous of total degree
`
X
|λ| = λi .
i=1

Notice that Sλ is unchanged if λ is extended by zeros:

S(λ,0) = Sλ .
The specialization Sλ (x1 , . . . , xn ) to one finite set of variables is called a Schur
polynomial, and Sλ (x1 , . . . , xn ; y1 , . . . , ym ) is called a double Schur polynomial.
sec:straight
1.3. Straightening law. For a, b ∈ Z and arbitrary integer sequences λ0 and λ00
we have
S(λ0 ,a,b,λ00 ) = S(λ0 ,b−1,a+1,λ00 ) .
In fact, the determinants defining these functions differ by interchanging two rows.
In particular, we have
S(λ0 ,a,a+1,λ00 ) = 0 .
A partition is a weakly decreasing sequence of non-negative integers, and we iden-
tify two partitions if they differ only by trailing zeros. For example, the sequences
(4, 3, 1) and (4, 3, 1, 0, 0) define the same partition.
It follows from the straightening law that any Schur function is given by
(
0 if λi − i = λj − j for some i 6= j;
Sλ =
±Sλe otherwise, where λ e is a partition.

In the second case, λ e is the unique partition for which the strictly decreasing se-
quence (λ e1 − 1, . . . , λ
e` − `) is a permutation of (λ1 − 1, . . . , λ` − `), and the sign of
Sλe is the sign of this permutation.
Exercise 1.1. The Schur functions Sλ indexed by partitions form a Z-basis of Λ.
 NOTES ON SCHUR FUNCTIONS 3

1.4. Skew Schur functions. Given two integer sequences λ, µ ∈ Z` , define the
skew Schur function Sλ/µ by


Sλ/µ = det Sλi −i−µj +j `×` .
For example,
S0+5−3 S1+5−1 S2+5−0
S(5,4,1)/(3,1,0) = S−1+4−3 S0+4−1 S1+4−0 .
S−2+1−3 S−1+1−1 S0+1−0
The function Sλ/µ is homogeneous of total degree
`
X `
X
|λ/µ| = |λ| − |µ| = λi − µj .
i=1 j=1

Notice that S(λ,0)/(µ,0) = Sλ/µ . We can therefore define skew Schur functions for
integer sequences λ and µ of different lengths by adding an appropriate number of
zeros:
Sλ/µ = S(λ,0,...,0)/(µ,0,...,0) .
The usual Schur function is Sλ = Sλ/0 .
Assume that λ and µ are partitions. We say that µ is contained in λ, written
µ ⊂ λ, if µi ≤ λi for all i. Notice that Sλ/µ is non-zero only if µ ⊂ λ:
eqn:skew_zero (1) Sλ/µ 6= 0 ⇒ µ ⊂ λ .
In fact, if µi > λi , then the i-th diagonal entry of the determinant defining Sλ/µ is
Sλi −µi = 0, and all entries south-west of this entry are also zero. It follows from
Theorem 1.4 below that the converse of (1) is also true.
The straightening law from Section 1.3 applies to both λ and µ. As a consequence
we have (
0 e 6⊂ λ;
if Sλ = 0 or Sµ = 0 or µ e
Sλ/µ =
±Sλ/e
e µ otherwise,

where λe and µe denote the partitions obtained from the straightening law applied
to Sλ and Sµ .

1.5. Young diagrams. A partition λ can be identified with its Young diagram
of boxes, which has λi boxes in row i. The row number i increases from top to
bottom, and the rows of boxes are left-justified. For example:

(7, 5, 5, 3, 1) =

The inclusion relation µ ⊂ λ means that the Young diagram of µ is contained in the
Young diagram of λ. When µ ⊂ λ, we let λ/µ denote the skew diagram of boxes in
the Young diagram of λ that are outside the Young diagram of µ. For example:

(7, 5, 5, 3, 1)/(3, 3, 1) =
 4 ANDERS SKOVSTED BUCH

A horizontal strip is a skew diagram with at most one box in each column:

A vertical strip is a skew diagram with at most one box in each row:

A skew diagram is called a rim if it is a union of a horizontal strip and a vertical

strip. Equivalently, the diagram contains no 2 × 2 squares. For example:

1.6. Expansions of Schur polynomials. In this section we let a = (a1 , . . . , an )

and b = (b1 , . . . , bm ) be finite sets of elements of a commutative ring R, and c ∈ R
denotes a single element. We first prove a basic formula for the expansion of a
double skew Schur polynomial defined by arbitrary integer sequences.
lemma:expand_seq Lemma 1.2. For any integer sequences λ, µ ∈ Z` , we have
X
Sλ/µ (a ; b, c) = (−c)|ε| Sλ/µ+ε (a, b) ,
ε∈{0,1}`

where the sum is over all sequences ε = (ε1 , . . . , ` ) with i ∈ {0, 1}.
Proof. Set h0p = Sp (a ; b, c) and hp = Sp (a; b) for p ∈ Z. Since the definition of
double complete symmetric functions implies that
X X
h0p tp = (1 − c t) hp tp ,
p p

we obtain
h0p = hp − c hp−1 .
The j-th column of the determinant defining Sλ/µ (a; b, c) is therefore given by
 0
hλ1 −1−µj +j
    
hλ1 −1−µj +j hλ1 −1−µj −1+j
h0λ −2−µ +j  hλ2 −2−µj +j  hλ2 −2−µj −1+j 
 2 j
 =   − c .
    
 .. .. ..
 .   .   . 
h0λ` −`−µj +j hλ` −`−µj +j hλ` −`−µj −1+j
The first vector on the right hand side is the j-th column of the determinant defining
Sλ/µ+ε (a; b) when εj = 0, and the vector multiplied to c is the j-th column in the
determinant defining Sλ/µ+ε (a; b) when εj = 1. The lemma now follows because
determinants are multilinear functions of column vectors. 
 NOTES ON SCHUR FUNCTIONS 5

When the sequence µ is a partition, the expansion of Lemma 1.2 can be inter-
preted in terms of adding vertical strips to µ.
prop:expand_part Proposition 1.3. Let λ, µ ∈ Z` and assume that µ is a partition. Then,
X
eqn:expand_vert (2) Sλ/µ (a ; b, c) = (−c)|ν/µ| Sλ/ν (a; b) ,
ν/µ vertical strip

where the sum is over all partitions ν containing µ, such that ν/µ is a vertical strip.
In addition,
X
eqn:expand_horiz (3) Sλ/µ (a, c ; b) = c|ν/µ| Sλ/ν (a; b) ,
ν/µ horizontal strip

where the sum is over all partitions ν containing µ, such that ν/µ is a horizontal
strip.
Proof. If µ ∈ Z` is a partition and κ ∈ {0, 1}` , then it follows from the straightening
law that Sλ/µ+κ is non-zero only if ν = µ + κ is a partition, and in this case ν/µ is
a vertical strip. Identity (2) therefore follows from Lemma 1.2.
Using (2), the right hand side of (3) is equal to
X X
c|ν/µ| Sλ/ν (a, c ; b, c) = c|ν/µ| (−c)|π/ν| Sλ/π (a, c ; b)
ν/µ horiz. µ⊂ν⊂π
ν/µ horiz.
π/ν vert.
 
 
X  X
c|π/µ| Sλ/π (a, b ; y)  |ν/µ| 

=  (−1) .
π ν: µ⊂ν⊂π 
ν/µ horiz.
π/ν vert.

The last two expressions are sums over partitions ν and π for which µ ⊂ ν ⊂ π,
ν/µ is a horizontal strip, and π/ν is a vertical strip.
It suffices to show that, if µ ⊂ π are partitions, then
X
eqn:cancel (4) (−1)|ν/µ| = δµ,π .
ν: µ⊂ν⊂π
ν/µ horiz.
π/ν vert.

If the sum is not empty, then π/µ must be a rim. Further, if ν satisfies the condition
of the sum, then any box of π/µ located immediately left of another box in π/µ
must be contained in ν, while any box of π/µ located immediately below another
box must be outside ν. If π/µ 6= ∅, then the North-East box of π/µ can be freely
added to or removed from ν. Since such a change to ν switches the sign of (−1)|ν/µ| ,
we deduce that (4) vanishes, as required.
?
π
ν ν π
π/µ =
ν ν ν π
π
π

 6 ANDERS SKOVSTED BUCH

1.7. Tableaux. Let x = (x1 , . . . , xn ) and y = (y1 , . . . , ym ) be two sets of distinct

commuting variables, and choose a total order on the union
x ∪ y = {x1 , . . . , xn , y1 , . . . , ym } .
Let λ/µ be a skew diagram. Define a bitableau of shape λ/µ labeled by (x; y) to be
a labeling T of the boxes in λ/µ with variables from x ∪ y, such that the following
conditions are satisfied:
• The labels of the boxes in each row of λ/µ are weakly increasing from left
to right with respect to the total order on x ∪ y.
• The labels of the boxes in each column of λ/µ are weakly increasing from
top to bottom with respect to the total order on x ∪ y.
• Given any variable xi from x, the set of boxes of λ/µ labeled by xi is a
horizontal strip.
• Given any variable yj from y, the set of boxes of λ/µ labeled by yj is a
vertical strip.
Any box of λ/µ will also be considered a box of T , and the label of a box will
be called the variable contained in the box. Let weight(T ) be the product of the
variables in all boxes of T , and set (−1)T = (−1)k , where k is the number of boxes
in T containing variables from y.
thm:bitableau Theorem 1.4. For any partitions µ ⊂ λ we have
X
Sλ/µ (x; y) = (−1)T weight(T ) ,
T
where the sum is over all bitableaux T of shape λ/µ labeled by (x; y), relative to any
chosen total order on x ∪ y.
Proof. This follows from Proposition 1.3 by induction on the number of variables.
For the induction step we use (2) if the smallest variable c in x ∪ y is from y, while
we use (3) if c is from x. 
Example 1.5. Let x = (x1 , x2 ) and y = (y1 ), and order these variables by x1 <
y1 < x2 . The bitableaux of shape (2, 1) labeled by (x; y) are:
x1 x1 x1 x1 x1 y1 x1 y1 x1 x2 x1 x2 y1 x2 y1 x2
y1 x2 y1 x2 y1 x2 y1 x2
We obtain
S(2,1) (x1 , x2 ; y1 ) = −x21 y1 + x21 x2 + x1 y12 − 2 x1 y1 x2 + x1 x22 + y12 x2 − y1 x22 .
1.8. Consequences of the tableau formula.
Corollary 1.6. Let µ ⊂ λ be partitions, and let x, y, and z be three sets of variables.
Then, X
Sλ/µ (x; y) = Sν/µ (x; z) Sλ/ν (z; y) .
ν: µ⊂ν⊂λ

Proof. We may assume that x and y are disjoint. Let z 0 and z 00 be arbitrary
(disjoint) sets of variables, and choose a total order on x ∪ y ∪ z 0 ∪ z 00 such that all
variables from x ∪ z 0 are smaller than all variables from y ∪ z 00 . Then Theorem 1.4
implies that
X
Sλ/µ (x, z 00 ; z 0 , y) = Sν/µ (x; z 0 ) Sλ/ν (z 00 ; y) .
ν: µ⊂ν⊂λ
 NOTES ON SCHUR FUNCTIONS 7

The result follows by setting z 0 = z 00 = z. 

Given a partition λ, the conjugate partition λT is obtained by interchanging rows
and columns in the Young diagram of λ. For example:
T
=

Corollary 1.7. For partitions µ ⊂ λ and sets of variables x and y, we have

SλT /µT (x; y) = (−1)|λ/µ| Sλ/µ (y; x) .
Proof. This follows from Theorem 1.4 because the transpose of a bitableau of shape
λT /µT labeled by (x; y) is a bitableau of shape λ/µ labeled by (y; x). 
Example 1.8. For any partition λ we have



Sλ (x) = det hλi +j−i (x) = det eλTi +j−i (x) .
Special cases include



hp = det e1+j−i p×p
and ep = det h1+j−i p×p
.
cor:vanish Corollary 1.9. Let λ be a partition. If λn+1 ≥ m + 1, then
Sλ (x1 , . . . , xn ; y1 , . . . , ym ) = 0 .
Proof. This holds because there are no bitableaux of shape λ labeled by (x; y) when
x = (x1 , . . . , xn ) and y = (y1 , . . . , ym ). For example, suppose T is such a bitableau,
subject to the ordering
x1 < x2 < · · · < xn < y1 < y2 < . . . ym .
Since row n + 1 of T contains at least m + 1 boxes, and each variable from y must
occupy a vertical strip in T , the leftmost box of row n + 1 contains a variable from
x. Since each variable from x occupies a horizontal strip, this implies that the
first column of T must contain at least n + 1 distinct variables from x, which is
impossible. 
Let (m)n = (m, m, . . . , m) be the partition containing n copies of m.
Corollary 1.10. We have
n Y
Y m
S(m)n (x1 , . . . , xn ; y1 , . . . , ym ) = (xi − yj ) .
i=1 j=1

Proof. It follows from Corollary 1.9 and the super-symmetry property that the
Schur polynomial vanishes if we substitute ym = xn :
S(m)n (x1 , . . . , xn ; y1 , . . . , ym−1 , xn ) = S(m)n (x1 , . . . , xn−1 ; y1 , . . . , ym−1 ) = 0 .
It follows
Q that S(m)n (x1 , . . . , xn ; y1 , . . . , ym ) is divisible by xn − ym , hence divisible
by i,j (xi − yj ) by symmetry. Since this product has the same degree as S(m)n ,
we deduce that
Yn Ym
S(m)n (x1 , . . . , xn ; y1 , . . . , ym ) = c · (xi − yj )
i=1 j=1
Qn
for some constant c ∈ Z. Finally, the easy identity S(m)n (x1 , . . . , xn ) = i=1 xm
i
reveals that c = 1. 
 8 ANDERS SKOVSTED BUCH

Exercise 1.11 (Factorization formula). Let λ ∈ Zn and µ ∈ Z` , and assume that

λi ≥ 0 for all i. For x = (x1 , . . . , xn ) and y = (y1 , . . . , ym ) we have
S(m)n +λ,µ (x; y) = Sµ (0; y) S(m)n (x; y) Sλ (x) ,
n
where (m) + λ, µ = (m + λ1 , . . . , m + λn , µ1 , , . . . , µ` ).

References
fulton:young [Ful97] W. Fulton. Young tableaux, volume 35 of London Mathematical Society Student Texts.
Cambridge University Press, Cambridge, 1997. With applications to representation the-
ory and geometry.
macdonald:symmetric*1 [Mac95] I. G. Macdonald. Symmetric functions and Hall polynomials. Oxford Mathematical
Monographs. The Clarendon Press Oxford University Press, New York, second edition,
1995. With contributions by A. Zelevinsky, Oxford Science Publications.